You are on page 1of 17

Journal of Applied Psychology 2005, Vol. 90, No.

1, 77–93

Copyright 2005 by the American Psychological Association 0021-9010/05/$12.00 DOI: 10.1037/0021-9010.90.1.77

High-Performance Work Systems and Occupational Safety
Anthea Zacharatos and Julian Barling
Queen’s University

Roderick D. Iverson
Simon Fraser University

Two studies were conducted investigating the relationship between high-performance work systems (HPWS) and occupational safety. In Study 1, data were obtained from company human resource and safety directors across 138 organizations. LISREL VIII results showed that an HPWS was positively related to occupational safety at the organizational level. Study 2 used data from 189 front-line employees in 2 organizations. Trust in management and perceived safety climate were found to mediate the relationship between an HPWS and safety performance measured in terms of personal-safety orientation (i.e., safety knowledge, safety motivation, safety compliance, and safety initiative) and safety incidents (i.e., injuries requiring first aid and near misses). These 2 studies provide confirmation of the important role organizational factors play in ensuring worker safety.

Most workers in developed countries assume their organizations will take all necessary measures to ensure that they return home safely at the end of the work day, yet work-related injuries and deaths continue to occur at an alarming rate. In the United States, there were 6,026 fatal work injuries and approximately 3.8 million nonfatal injuries in 1998, resulting in an estimated 80 million production days lost for that year and almost 60 million days in future years (Bureau of Labor Statistics, 2000; United States Census Bureau, 2000). In 1999, there were 833 work-related fatalities in Canada, while 379,395 Canadian workers suffered injuries serious enough to be compensated either for wages lost due to time off from work or a permanent disability (Association of Workers’ Compensation Boards of Canada, 2000). These data illustrate the enormous cost of occupational injuries and fatalities for organizations in terms of production and for lives altered and lost by these work-related events. Traditionally, the most frequent method for managing occupational safety has been by taking a control-oriented approach to human resources (Barling & Hutchinson, 2000), one that assumes workers are motivated to exert only as much effort as is necessary for task completion. As such, it is management’s responsibility to use its legitimate authority to control employee behavior (Walton, 1985). In terms of occupational safety, the control-oriented approach emphasizes the use of rules to enforce behaviors and the use of punishment to increase rule compliance (Barling & Hutchinson, 2000). There has been a growing realization that human resources are better managed by high-commitment- (e.g., Walton, 1985) or
Anthea Zacharatos and Julian Barling, School of Business, Queen’s University, Kingston, Ontario, Canada; Roderick D. Iverson, Faculty of Business Administration, Simon Fraser University, Vancouver, British Columbia, Canada. We acknowledge assistance from the Industrial Accident Prevention Association and the Social Sciences and Humanities Research Council of Canada. Correspondence concerning this article should be addressed to Julian Barling, School of Business, Queen’s University, Goodes Hall, Kingston ONT K7L 3N6, Canada. E-mail: 77

high-involvement- (e.g., Lawler, 1996) oriented strategies. Wood (1999) noted that these approaches reflect a rejection of the traditional Taylorist model and a heightened focus on job-design theory. Rather than relying on compliance by means of rules, regulations, and monitoring to decrease costs and increase efficiency, high-commitment management creates conditions that encourage employees to identify with the goals of the organization and to exert effort to achieve them (Whitener, 2001). Similarly, highinvolvement management concentrates on empowering employees through increased information flows and devolution of decision making power, leading to greater productivity. More recently, the term high-performance work systems has been used to characterize these transformed workplaces. Although high-performance work systems encompass the high-commitment and involvement elements, they are also broader in scope by emphasizing the competitive advantage gained by such human resource practices.1 Way (2002) and Wood and Wall (2002) conceptualized high-performance work systems as a group of separate but interconnected human resource practices that together recruit, select, develop, motivate, and retain employees. Way (2002) suggested that this is achieved by ensuring that employees possess a broad range of superior skills and abilities that are used at work, which ensure that their organizations achieve “superior intermediate indicators of firm performance (i.e., those indicators over which the workforce has direct control) and sustainable competitive advantage” (p. 765). High-performance work systems assume employees are a primary source of competitive advantage that is difficult for others to imitate and that workers are capable of continuous improvement and will perform at higher levels if they are motivated to do so (Pfeffer, 1998a). This is achieved by encouraging practices such as participative decision making, providing high-quality training, and sharing information. By treating workers with respect and as capable and intelligent individuals, organizations will find that
Wood, de Menezes, and Lasaosa (2003) noted that researchers continue to use these three forms of management interchangeably.

Moore. Employment security encourages a long-term perspective and represents an investment of time and resources in employees. Bielema. In the present study. 1992. & Kelloway. Employment security differentiated highaccident-rate companies from low-accident-rate companies (Smith. high-quality work. 1997. although it is beyond the scope of this article to discuss the universalistic. 1980). or customer complaints (Griffiths. Ichniowski et al. 2004. which will result in improved performance (Walton. The way in which hiring is used to achieve occupational safety is usually consistent with a control orientation. we expected to find greater safety behavior. Way. causal inference from some of these studies is compromised by statistical and methodological issues such as the use of cross-sectional data and the absence of true control groups. it is generally agreed that they should be multiple and mutually reinforcing (Cappelli & Neumark. Trust in management will also result from employment security. 1995.78 ZACHARATOS. greater employee productivity in terms of sales per employee. emotional maturity. Wood & Wall. Furthermore. & Prennushi. Although there is some debate regarding the number of human resource practices that constitute a high-performance work system. Therefore. This assertion follows the argument often found in the literature that safety should be considered a performance variable much like production. as well as three additional practices that we deem to be equally important (transformational leadership. Becker and Huselid (1998) concluded that it is theoretically appropriate to consider a high-performance work system as a single system and that “the overwhelming preference in the literature has been for a unitary index that contains a set (though not always the same set) of theoretically appropriate HRM [human resource management] practices derived from prior work” (p. MacDuffie. 1993. not to replicate and test other high-performance work systems. Empirical data also support a link between employment security and occupational safety. J. B. These include Pfeffer’s (1998a) seven factors (employment security.. Kalimo. 2002). 1995. 2002). 1998. 1995). 63). and trustworthiness (e. and we hypothesize specifically that a high-performance work system will improve workplace safety by increasing employee trust in management and perceived safety climate. Employees of organizations with more traditional approaches to human resource management produced less than their peers. Employees view human resource practices and trustworthiness of management as indicative of the organization’s commitment to them. Borofsky. employees reciprocate with appropriate attitudes and behaviors. turnover rates) were significantly better in minimills operating under a commitment-oriented system than in minimills managed in a control-oriented fashion. employment security. and both market-based and accounting-based measures of corporate performance. Cohen. BARLING. Typically. Barling. 1997). One of the most basic ways in which organizations can improve their performance is by ensuring employment security (Pfeffer. and contingent compensation). West. 1997). We now describe these practices individually. Huselid. quality control. which is desirable to the extent that trust in management is associated with organizational performance (McAllister. we propose a set of 10 practices that have been theoretically and empirically associated with occupational safety. Jones. 2003. AND IVERSON workers will be more committed to the organization and more trusting of management. Huselid.g. J. 2001).. & Wall. 1998a). & Greenberg. sales. Huselid (1995) examined 958 publicly traded companies and reported that high-performance work systems were associated with significantly lower turnover rates. crosssectional and longitudinal data showed that job insecurity was related to lower safety motivation and safety compliance and greater numbers of workplace injuries in a sample of foodprocessing-plant employees (Probst & Brubaker. Arthur. which would be reciprocated in terms of loyalty to the organization (Tsui. Wheatley. Pfeffer (1998b) stated that it is important for organizations to measure indicators that are important to their particular business and that successful companies often have performance standards that are unique and go beyond typical financial reporting measures. In a study of 30 minimills. 1995). (1997) found that steel output was greater when organizations had in place practices such as incentive pay. selective hiring. & Iverson. Selective Hiring Selective hiring focuses on the fit between employees and their work environment. for a comprehensive review). profits. Zohar. with its reliance on the “selective exclusion” of high-risk employees. 2001. We extend current research and predict that high-performance work systems will also influence occupational safety. B. In turn. 2001).g. and measurement of management practices) in predicting occupational safety (Barling. 1995. 1985. Ichniowski. 1996). personality questionnaires are used to prescreen potential employees on the basis of current or former drug addiction or alcoholism. 1985. Consistent with a control orientation. Patterson. & Cleveland. & Tripoli. Porter. Cohen. Pearce. 1978. Numerous studies now provide empirical support for the superiority of high-performance work systems for employee and organization-level performance (e. Similarly. We argue that high-performance work systems can be applied to improving workplace safety just as well as firm economic performance. Likewise. 1994. elaborate screening of new employees. Shaw. . Loughlin. Arthur (1994) found that performance quantity (in terms of labor hours) and performance quality (as measured by scrap rates. and off-the-job training. 1991). and feelings of greater job insecurity were positively associated with the actual number of injuries and days of work missed because of an injury (Grunberg. and con- figurational approaches of high-performance work systems (see Becker & Huselid. Kivima ¨ ki. Kelloway. information sharing. This would suggest that employees who have a poor fit or who are mismatched with their environment display a greater propensity to be injured (Iverson & Erwin. & Salminen. reduced status distinctions. Whitener (2001) proposed a social-exchange framework to explain this relationship. our aim is to develop a system of high-performance work practices. & Hoffman. on the basis of Pfeffer’s (1998a) framework. extensive training. 1997). Employment Security Employment security refers to the extent to which an organization provides stable employment for employees. contingency. In addition. problemsolving teams. following which we consider how they together compose a high-performance work system. this approach rests on the assumption that work-related injuries are primarily the fault of employees. teams and decentralized decision making. flexible job design.

were the best predictors of the propensity to comply with safety rules.. Driskell. 1977). Simard and Marchand’s (1997) study of 97 manufacturing plants showed that the quality of supervisor– employee relationships. and increased sharing of ideas (Trist et al. In contrast. Similarly. Similar results emerged when miners working in autonomous work groups attributed their improved safety performance to their common goals. workers may be inclined to hold management responsible for workplace injuries. Trust in management will also be enhanced because of employees’ perceiving the organization as valuing and caring about them (Eisenberger. In a highperformance work system. Members of a flight crew performed more effectively as a group in dangerous situations than as a hierarchy with the captain at the top of the chain of command. Although each was described identically in terms of age. Milanovich. The participants in their study held higher general and specific expectations for pilots than copilots. 1977). both Cohen (1977) and Smith et al. there is a tendency for conclusions about injury causation to be biased. Employees must also be empowered to use new skills following training (S. Tjosvold (1990) showed that with more cooperative goals. teamwork and decentralized decision making foster familiarity and demand greater cohesion. Roberts. K. they are more likely to see their own safety as being dependent one on the other and to feel an increased responsibility for joint safety. 2001). 2001). Parker. The researchers showed that this status difference was critical. who argued from a human-capital perspective that training increases the problem-solving skills of employees. Essentially. Extensive Training Occupational safety training is likely the most researched issue and practiced technique in safety management. These results are important because organizational commitment predicts work performance in general (Meyer & Allen. 2003). safety infractions increased.OCCUPATIONAL SAFETY 79 Organizations committed to occupational safety will attend closely to how they hire new personnel and will incorporate the value of occupational safety into their employee-selection processes to achieve a better fit. To study the effects of status distinctions. it is perceived to be in top management’s best interest to blame front-line workers for safety infractions. and harm motivation and performance (Pfeffer. mental status. Parker. as illustrated by a study of naval trainees that found that the extent to which training was perceived to have met trainees’ initial expectations. when managers and employees see these status barriers being dismantled. employees working in autonomous teams in a mine experienced fewer accidents than did employees who worked individually (Trist. Working in teams should also promote safety because it causes individuals to feel more responsible for their own and each others’ safety. as well as cohesion with the work group. belongingness to a group and personal control predicted workers’ propensity to actively care for coworker safety. This view is consistent with Osterman (1995). Wall. Teamwork and decentralized decision making should also enhance occupational safety because it provides those people who are more familiar with the situation greater opportunities for control. 1991). and Gilmore (1996) found that sense of . 1997) and safe working in particular (S. & Brown. A study of injured construction employees illustrates another aspect of status distinctions. the safest teams in a chemical plant were those with the most control over varied aspects of their work (Hechanova-Alampay & Beehr. Although research examining the relationship between selection practices and occupational safety is sparse. & Jackson. increased communication between members. and employees who receive safety training suffer fewer work-related injuries than their untrained counterparts (Colligan & Cohen. training allows employees to acquire greater competencies to control their work. Goodman and Garber’s (1988) study of safety in underground coal mining found that as the familiarity between members of a working dyad decreased. (2003). Reduced Status Distinctions Status distinctions in organizations are ubiquitous. Stout. In line with attribution theory. & Turner. Baltz. Status distinctions also have the negative effect of reducing the familiarity between top management and shop floor employees. the other as a copilot. presumably because team members were motivated to contribute their ideas by the belief that they must work together in achieving their goals (Tjosvold. (1978) found that lower injury rate companies had more elaborate selection procedures than organizations with higher injury rates. in which placing blame for events is so inherent. Indeed. Axtell. Last. Susman. & DavisLaMastro. Gillen. and the learning of academic content all predicted subsequent organizational commitment (Tannenbaum. create unwanted barriers between people that breed resentment. one was described as a pilot. 1990). all members of a flight crew felt responsible for safety and handled critical incidences more effectively. & Cannon-Bowers. teams will enhance occupational safety when they promote the sharing of ideas that result in better solutions. Training also exerts indirect effects on safety. Kirch. training must extend beyond the mere provision of knowledge related to how to do one’s job safely. and health. 1997). each employee from the shop floor to top management should feel that they can contribute to diverse Self-Managed Teams and Decentralized Decision Making There are several reasons why teamwork and decentralized decision making should benefit employee performance and safety performance. 1998a). how satisfied trainees were with the training. Salas. and Salas (1998) asked 30 recently qualified military aviators to rate their expectations for two aviators. DeJoy (1994) argued that in the case of occupational-safety management. K. and Vaccaro (2002) observed that union employees were more likely to perceive their supervisors as caring about their safety compared with nonunion employees. DeJoy (1994) suggested that one of the ways of confronting this is to reduce the distances between employers and employees through increased exposure to each other. experience. First. Gassel. Fasolo. 2001). To be maximally effective. As noted by Barling et al. 1990). and manufacturing teams with greater authority over their work experienced fewer workrelated injuries (Kaminski. Mathieu. leading to them performing their jobs more safely. Involving teams in selecting future members could indirectly serve to improve safety levels and requiring applicants to go through several rounds of interviews in which the organization’s values are conveyed might also enhance occupational safety. Geller.

& Fitch. The number of days lost from work because of injuries. meaningless work. Olson. feedback. they were more highly committed to following safety procedures and practices. there is widespread concern that such approaches rely on managerial control (Walker. 1998). Safety infractions increase during periods of economic growth.. through token reinforcement) is effective in reducing occupational injuries (e.80 ZACHARATOS. would show more trust in management. 1979.. 2002). 2003). 2. and employee control (e. Smith et al. High-Quality Work It would avail little to implement a high-performance work system if employees were left with boring. a task requiring strong leadership. we argue that high-quality work should be a critical component of a high-performance system. Overall. Similarly. Transformational leadership is (a) positively associated with safety initiative (O’Dea & Flin. Collectively. Karasek. if safety is considered a key performance indicator by an organization. Fox. 1998. these studies support the idea that transformational leadership plays an important role in occupational safety. and the behaviors under consideration are typically highly specific. it is the organizational leader who chooses the nature of the system to be implemented who helps to ensure successful implementation (Pfeffer. 1998a) and who helps to choose the outcomes that the organization values.. the effectiveness of transformational leadership is supported in a number of contexts (Bass. skill use). and. Organizations with better safety programs (Zohar. There are data showing that paying people contingent on their safety performance (e. BARLING. 1977. Information Sharing Just as information sharing across organizational levels is critical for high performance in general. There are also secondary benefits of information sharing for occupational safety. 1987) and dimensions of job quality (i. 1980. Third. each of the four transformational-leadership factors lend themselves to the task of enhancing safety performance. employees should be rewarded for their efforts to improve safety. 1999). 1989).e. several problems emerge with this approach: Its long-term effects are not fully understood. and by explicitly choosing which behaviors are to be rewarded. research shows that transformational leadership can be taught to managers (Barling. 1996). organizations signal unambiguously which behaviors are valued. More consistent with Pfeffer’s (1998a) proposition is the idea that (a) employees are compensated for safety performance at the group level and (b) compensation be provided for behaviors that extend beyond an individual’s personal safety. doing so means going against the tide. 1998). Pine. (b) negatively associated with minor injury rate (Zohar. 1996). and (c) indirectly associated with injuries (Barling et al. employees would appreciate those instances when management took safety seriously and. Parker & Wall. 2000) and the priority assigned to safety (Zohar. Levine. AND IVERSON aspects of the organization..g. Kochan. the number of lost-time injuries. we emphasize appropriate workload. employees will have greater trust in management. 2002). 1978) were characterized by more open discussion between management and employees. 3. The idea that work overload might be linked to occupational safety is not new. Moreover. K. task significance.g. presumably because the need for greater production to meet demand results in an increase in work pace . therefore. Transformational Leadership There are several reasons why transformational leadership would be an appropriate leadership model for enhancing occupational safety and why it should be considered integral to a highperformance work system. and the costs of accidents and injuries were substantially reduced.. a well-designed job will ensure that employees are engaged intellectually and emotionally. 4.. when employees felt comfortable discussing safety-related issues with their supervisors. & Strauss. Nonetheless. Although there are numerous conceptualizations of job quality (Hackman & Oldham.. Weber. As such. 2002). 1980) and safety records (Cohen. Clarke (1999) found that both management and employees underestimated how important safety was to the other group and argued that with greater information sharing. First. Warr. these remained lower over several years. Intellectually stimulating leaders help followers think about safety and develop new ways to achieve high safety levels. Compensation Contingent on Safe Performance Well-paid employees feel valued by the organization. Barling et al. we argue that it is equally critical for optimal safety performance. As most organizations do not make use of systems of high-performance work practices (Ichniowski. 1. There are empirical data to support the link between transformational leadership and occupational safety. Tokens were also provided for making suggestions to improve safety and for taking unusual actions to prevent an injury or an accident. Employees were given tokens redeemable for goods at the end of the month if neither themselves nor a member of their workgroup had been involved in a safety-related incident in the preceding month. in turn. Haynes. & Kelloway. Where managers do share information. Hopkins. Individualized consideration would be evident through leaders’ real concern about their followers’ safety at work (Barling et al. Leaders high in idealized influence would convey the value of safety through their personal behaviors. limiting the possibilities that what is learned will generalize across situations (McAfee & Winn. and several studies support the role of information sharing in occupational safety. Furthermore. Second. 2000). and Anger (1987) found support for this notion by studying employees from an open-pit mine. 2002) and microaccidents (Zohar.g.. S. which resulted in the lower occurrence of workplace injuries (Hofmann & Morgeson. 1982). it would not be possible to work safely without full information about all aspects of one’s job. Arguably. we argue that reduced status distinctions are critical for occupational safety. Those high in inspirational motivation would convince their followers that they could attain levels of safety not previously considered possible. Last. Indeed. role clarity.

showing that human resource management practices are more likely to yield positive effects for the firm when they are introduced as part of a coherent system.g. Accordingly. Responses were received from the human resource directors of 147 organizations. Parker et al. For instance. Arthur. Frone (1998) found that job boredom was associated with adolescents’ workplace injuries. 138 of whom provided surveys with usable data (response rate ϭ 9. a further purpose of this study was to determine the underlying factor structure of the 10 human resource practices. Study 1 Method The purpose of this first study was to determine whether a relationship exists between the high-performance work system described and occupational safety at the organizational level. The organizations in the final sample belonged to a wide range of industries. For instance.. confidential information with individuals who may subsequently work for the competition. 25).. When managers feel hindered by an unusually heavy workload. we investigated the relationship between human resource management practices and safety performance at the organization level.” Dyer and Reeves (1995) also concluded that these systems “seem to be superior to any of the individual human resource activities of which they are composed” (p. 2001. Given the conceptual overlap between each of the high-performance work practices. 1999). Measurement of Management Practices The safety behaviors that organizations typically measure— namely. Autonomy (S. For 35 of the items. Dembe. Hemingway and Smith’s (1999) study of 252 nurses employed in four different hospitals showed that nurses’ role ambiguity was significantly associated with the number of injuries they experienced. 1998). Job autonomy enhanced employees’ commitment to the organization. which Wheatley (1997) argued. The organizations had been in existence for an average of 43 years (SD ϭ 31. automotive. group cohesion and cooperation.OCCUPATIONAL SAFETY 81 (Conway & Svenson. 1998a). 2002). 1996.75). K. Purpose of the Current Research In the first study. 1999. compliance with government regulations. Role clarity is a further dimension of job quality that is important for occupational safety. positive employee attitudes. Measuring the number of safety incidents provides little insight into near misses or injuries not requiring medical attention that may occur with greater frequency (Hemingway & Smith. The second study focused on the employee level and explored mechanisms that mediate the relationship between the high-performance work system and safety performance.38%). Snyder. K. 1996). 1999) and the United Kingdom (Wood & Albanese. and a study of 71 transport drivers showed that attention problems predicted actual involvement in vehicular incidents (W. “employees in this organization are involved in the hiring of their peers”). safety is compromised (Baugher & Roberts.53). and construction. Kling’s (1995) study took the analysis of high-performance systems a step further. Simard and Marchand (1995) demonstrated that participative management predicted (a) the extent to which employees were proactively involved in their own safety (as opposed to compliance with safety regulations) and (b) two other factors that are critical for safety. Participants Surveys were sent to 1. Ramsay. yet it is important because boring tasks lead workers to seek stimulation from other sources (Fisher. relying more on contingent and contract workers. & Harley. past incidents and injuries. measuring management practices that increase employees’ levels of trust in management and perceived safety climate would be beneficial to the extent that these factors predict subsequent safety performance. The majority of workplaces had nonunionized front-line employees (57%).32) and with the organization an average of 8 years (SD ϭ 8. namely. Becker & Huselid. Responses to this type of . A System of High-Performance Practices Describing each practice separately in no way implies their independence. another 34% had unionized front-line employees. 6 –7 items per practice). 1999. & Himmelstein. 1998. including chemical. may see little benefit in expending scarce resources on selective hiring or extensive train- Materials High-performance management practices were measured with 63 items developed for the current study (i. Guthrie. Way. Approximately 95% of the employees worked full time.. 1990). Pfeffer. 1998) and job-decision latitude (Karasek & Theorell. 1995) support the interdependence of the high-performance practices. 2002) and encourages the under-reporting of safety infractions (Eisenberg & McDonald. The human resource professionals who responded to the survey had been in their positions an average of 5 years (SD ϭ 5. Barrett. What is needed to enhance safety in the long-term is a measurement approach that focuses on the proximal causes of safety incidents. 1990) are further elements of a high quality job and are associated consistently with work performance. organizations that are averse to providing employment security. Scholarios.e. we created a unitary index for our high-performance work systems (e. 668). Parker & Wall. which in turn affected their safety compliance with procedures (S. participants rated the extent to which they believed the practice existed in their organization (e.g. Hofmann & Stetzer. with an average of 515 employees (SD ϭ 828. instituting one management practice logically drives the need for the application of other practices (Becker & Gerhart. & Doverspike. (1996) reported that lost-time accident rates were lower in workplaces in which workers participated in decision making. Shannon et al. and provisions of collective agreements—are consistent with and foster a control-oriented climate. ing and will likely be less willing to share valuable.471 manufacturing organizations that were members of the Industrial Accident Prevention Association of Ontario. and physical and psychological wellbeing.88). We hypothesized that organizations applying high-performance management practices will experience lower numbers of lost-time injuries after controlling for relevant variables. rather than as single “best practices. Autonomy at work is also associated with enhanced safety. 1993). Zohar. 1988. Pransky. “providing employment security to our employees is a priority in this organization”. 2000. and the remaining 9% of workplaces had both unionized and nonunionized front-line employees.. “can never spell out the route to perfect safety” (p. Empirical data from the United States (Wood. Less attention has been focused on the effect of work underload on occupational safety. On the contrary. 2001).

cuts..41** .35 . . selective hiring.48** 3 4 5 6 7 8 9 10 .42** . 1978). (g) eye injuries.76 1 . the sum of scores on each item was calculated and analyses were conducted on these summed scale scores.01 3.02 Ϫ. we used an additive approach to create a unitary index..g. 3. the reliability coefficients for the subscales measuring each of the 10 high-performance management practices were calculated and those items that most negatively impacted the internal consistency of each scale were removed.g.g.28** .77** . They reported the identical factor solution between item-level data and the corresponding summed subscale data across five hospital samples. Results To conduct analyses on the varied forms of responses to the items measuring high-performance management practices. a format used previously in this type of research (e.02 Ϫ. BARLING.65 . 2001.05. the average item intercorrelations for the five factors were respectively .32.56 . AND IVERSON Table 1 Descriptive Statistics and Intercorrelations Between Measures of the Ten High-Performance Management Practices in Study 1 (N ϭ 138) Variable 1.g.97 Factor loading .27.56** . we conducted confirmatory 2 This questionnaire. .69** . 2000: (a) fractures. and 7..20* .28** . and Williams (1999) argued that. By focusing on these subjective and objective elements.. Comrey.54** .56** .07 2. factor analysis. Arthur.05 . Kraus. 10. 3.80 .35** . B. researchers have typically used a time frame of the past 12 months (e. for a review).66 Note.25** .35** . and . Guthrie.26** . Third. scratches and abrasions). (e) open wounds (i. Barling et al. 2001.72 . status distinctions. 2001). .02 . 5.27. 2002). Becker & Huselid.44** . Accordingly. White. Second.43** .44** .64 .30 4. we could better gauge the usage of each practice. the basic aim of our research was to assess whether the greater use of a “system” of high-performance management practices would improve workplace safety.46** . 2003). in italics. they are included in the analysis for several reasons.28** . df ϭ 31.e. given the number of items. it must be interpreted with the number of items in mind” (p. 1998.59 . Employment security Selective hiring Training Teams Status distinctions Information sharing Contingent compensation Transformational leadership Job quality Measurement M .58** . sprains. we used a multimeasurement approach to identify both the existence and extent of each human resource practice.05 3. question were on a 5-point Likert scale ranging from strongly disagree (1) to strongly agree (5).21. as well as all others used in this research.01. (c) bruising and crushing. 7. the number of items for each of the five variables was 5. Huselid.50** . and (h) concussions and other head injuries. . as human resource managers frequently have access to such statistics from organizational records. Sanchez. Injury Rates Respondents reported the number of lost-time injuries and total number of days lost due to each of eight specific types of injuries between January 1 and December 31.52** . lacerations. 3 As Cortina (1993) noted “alpha is very much a function of the number of items in a scale .81 . First.23 2. “this type of subscale aggregation is likely to provide more reliable indicators than item-level data” (p. Becker and Huselid (1998) further acknowledged that one or a combination of practices that are more heavily emphasized would have the same index value as a wider range of modestly emphasized practices.01 ..57 . Drasgow and Kanfer (1985) have provided empirical support for the subscale aggregation approach.91 . . 3. That was one criterion. 1993).05 SD 3. Becker and Huselid (1998) noted that this approach is appropriate because it suggests that (a) a unitary human resource system is a strategic asset and (b) there are many ways to increase the organization’s value.58** . First.58** . consistent with our conceptualization of high-performance work systems. contingent compensation. 469).47** . Finally. and punctures). is available from the authors on request.75 .3 To investigate the underlying factor structure of the 10 highperformance management practices. Guthrie.52** . these scales are retained in the study (Cortina. 1995).78 . 9.01 . 8.33** .e.. 5. Subsequently. .21* .45** . This allows researchers to address problems such as distribution violations and statistical power issues associated with item-level data (e.67** .55** . and item intercorrelations and the fact that the scales were developed by the researchers.51** .12 4.76 . Descriptive statistics and intercorrelations between the 10 high-performance management practices appear in Table 1. respondents estimated the percentage of employees in their organization to which a statement describing a high-commitment management practice applied (e. Although the scales of employment security. respectively.22. and strains. 2. For each subscale. 1992. The other criterion was that the resulting subscale had to be consistent with our conceptualization of the construct. This measurement strategy maximizes several critical factors.01 Ϫ.2 Two further items inquired about rounds of layoffs and the number of hours of nonmandatory training received by employees in the past 12 months.17 .30** 2 . * p Ͻ . p Ͻ .70.g. 102).. to address the limitations of previous survey research on high-performance work systems (see Godard. J.53 . (b) dislocations. (d) superficial wounds (i. congruent with previous approaches in the literature (e.52 . For 26 of the remaining questions. Second.63 .23** .23 4. we standardized all item scores (Way. ** p Ͻ . 6.77 .83 . (f) burns and scalds.44** .31** .01).38 4. “What percentage of front-line employees received training beyond that mandated by government regulations in the last 12 months?”).67** .29** .58** .52** .86.70** . Respondents were also asked to report the number of fatalities that occurred in the same time period. The reliability of the lost-time-injury data was checked by correlating responses from the 32 companies where responses were received from both human resource and safety managers (r ϭ . Therefore.01 Ϫ.08 3. and measurement displayed alphas lower than .47** . Scale reliabilities can be found on the diagonal.82 ZACHARATOS.48** .. factor analysis using principal components established unidimensionality of the factors. 4.49** .

08 . 1995). There is debate as to whether subsidiaries have similar or different human resource practices as their parent organizations (Martell & Carroll. of employees High-performance work system Total lost-time injuries a — Ϫ. 1997). 3..88 0. fatalities were not analyzed further.21 2 3 4 5 6 7 Subsidiary Private–publicb Age of organization Unionizedc No..59. 1999.09 Ϫ. accounting for a significant 8% of the variance above that of the control variables. three important findings emerged.64. Wood & Albanese.02 . On the other hand.49 0. ** p Ͻ . Wood. selective hiring. larger firms are also more likely to be under greater scrutiny by health and safety inspectors. The results of the LISREL analysis are presented in Table 3. 1 ϭ private. Bentler.15 Ϫ. whereas private and older organizations are argued to be associated with more sophisticated human resource practices (Guthrie. Third. 1986).22 17. The variance– covariance matrix was used for input. 1995). the reliability of our survey instrument would be improved by better psychometric measures of our human resource practices of employment security. 1993). In the analysis. unions may enforce better reporting systems leading to the impression of higher injury rates (Kelloway.97. The hypothesized one-factor model demonstrated a significantly better fit than the null model ⌬␹2(10.03 .54 0.04 Ϫ. high-performance work systems were related to fewer lost-time injuries.45** — Ϫ. 2004. private/public. 1993) arising from our sample size. Because of the problems associated with the nonnormal sample distribution of total number of lost-time injuries (i. contingent Table 2 Descriptive Statistics and Intercorrelations Between Measures in Study 1 (N ϭ 138) Variable 1.89). 2 ϭ public. Browne & Cudeck. 1 ϭ unionized. Second. unionized. F(1. However.35 to . and a root-mean-square error of approximation (RMSEA. J. Second. 7. A very small proportion of organizations reported occupational fatalities. The variable measuring high-performance management practices was entered into the equation in Step 2.01 Ϫ.05 .001. 2003).26** — 1 ϭ subsidiary. status distinctions. .01.93. In addition. the cross-sectional nature of the data makes any inferences about causality somewhat risky.00 1. 1990) of .59 1 — Ϫ. p Ͻ . we would expect a negative association with lost-time injuries (Sprigener & Hodson. 1999).e. a manifest-variables model was used (L. we control for several organizational differences (subsidiary. the single-source nature of the data introduces the possibility of measurement error (Wright et al. the results of this study replicate earlier research showing that high-performance work practices at the organizational level are associated with employee and firm performance and go beyond these traditional outcomes by demonstrating that a high-performance work system also affects occupational safety.75 515. the CFA conducted as part of this study should provide more meaningful results than those of previous studies.14 41. as a result of which. 1996.75 25. several limitations of this study need to be confronted. 6.61 828. Nevertheless. where the latent-to-manifest parameter for high-performance work system was fixed to the square root of the reliability (internal consistency coefficients) and the value of one minus the reliability multiplied by high-performance work systems’ variance was used to calculate residuals (Carlson & Perrewe ` . the 10 high-performance management practices composed a single underlying construct—a high-performance work system. 131) ϭ 15. Genn. 2001).10 .14 — Ϫ.05. 4.e. because our measure of a high-performance work system follows the recommendations of Becker and Huselid (1998) and includes multi-item measures for each practice. Type II [␤] error.91.OCCUPATIONAL SAFETY 83 factor analysis (CFA) using LISREL VIII (Jo ¨ reskog & So ¨ rbom.12 .05 .. with a goodness-of-fit index (GFI) of .35** Ϫ. 1987). p Ͻ . 1993) of . Fourth. we combined all 10 subscales to construct a single index measuring the high-performance work system (␣ ϭ . Descriptive statistics and intercorrelations of all variables are presented in Table 2. 2 ϭ not a subsidiary. 1992) on the summed scale scores (Sanchez et al. 2001).11 — . First. * p Ͻ . we censored the variable. identifying a substantial predictor of safety performance. Third. an organizational representative provided the data regarding the existence of human resource practices in each respective company as well as the firm’s overall safety performance. 1988).07 . a M 0. Discussion In this study. 2. Because the 10 practices consisted of a single factor. As predicted. 5.20* c — Ϫ. skewed and truncated). age of organization.68 b SD 0. Saris & Satorra.01 Ϫ.01 Ϫ. 1999). As unions are a key social mechanism in establishing safety minimums and standards.06. high-performance management practices accounted for 8% of the variance in lost-time injuries measured at the organizational level. Finally. leading to enhanced safety consciousness and lower injuries (Fenn & Ashby. Williams & Hazer.001 (Bagozzi & Yi. a normed comparative fit index (CFI. First. N ϭ 138) ϭ 624.63 1.33 31. larger organizations (as measured by the total number of employees) could increase the propensity of injuries because of increased bureaucracy and poorer coordination and communication (Hopkins & Palser. Of the control variables in Step 1. and the total number of employees were entered into the equation in Step 1). 2 ϭ nonunionized. This supports earlier findings of other systems of high-performance practices at the organization level (Huselid & Becker. Examination of the parameter estimates (factor loadings) of the one-factor model were all significant ( p Ͻ . only subsidiary status and the total number of employees were related to lost-time injuries.05) and ranged from .12 43. Because of the potential power problems (i.

yet such measures may underestimate the number of actual injuries (Eisenberg & McDonald. we focus on employee perceptions of high-performance practices in this study.28** .g.25** Step 2 Ϫ. Kramer (1999) reported that trust results in spontaneous sociability. Safety climate has also been related to safety compliance and employees’ participation in safety-related activities (Neal. Seventh. 2003) and transformational leadership and follower performance (e. to look out for the safety of fellow employees. work toward common goals. 2 ϭ not a subsidiary. Because organizational records can be of questionable validity (see Eisenberg & McDonald.03 . some caution needs to be noted with the measure. Last.15. Sixth. We assessed three aspects of safety incidents. (a) microaccidents—those injuries that require a visit to the infirmary but do not require time off of work (e. Standardized LISREL coefficients are reported. no indication is provided as to how high-performance work systems influence occupational safety. with that person typically being a human resource director or other company official. the results of this first study show that high-performance work systems are associated with occupational safety. unionized. Jung & Avolio. of employees Step 2 High-performance work system Total R2 ⌬R2 a Step 1 Ϫ. we looked to different safety outcomes as measures of safety performance. 2 ϭ public. and information-sharing. Zohar. Schalm and Kelloway (2001) found a weighted average correlation between effect size and response rate in self-report survey research (N ϭ 177) to be . with a nonsignificant population variance estimate of . 2 ϭ nonunionized. Many of the practices associated with high-performance work systems are hypothesized to result in an individual’s positive perceptions of safety climate..06 . Last. 2002.02. 2000). Although low. For example. especially during turbulent and unstable times in the organization. In turn. when management is seen to offer extensive training because it is committed to employee safety. In addition. Pransky et al. 1980). “nonresponse is not likely to result in substantial bias in the results” (p. 2000.84 ZACHARATOS.17* Ϫ. perceived safety climate is a proximal predictor of safety behaviors. They concluded. In a review of the trust literature. Pransky et al. ** p Ͻ . Second. Loughlin. & Sherman. In doing so. 1ϭ compensation. we extended Study 1 in several ways. * p Ͻ ... AND IVERSON Table 3 Study 1: LISREL Results Predicting Total Lost-Time Injuries ␤ Predictor variable Step 1 Subsidiarya Private–publicb Age of organization Unionizedc Total no.09 . perceived safety climate is enhanced.03 . 2000). we hypothesize that any such effects are indirect and will be mediated by perceived safety climate and trust in management (see Figure 1). 2000. we addressed some of these issues by investigating the link between the high-performance work system and occupational safety at the employee level. 2000). procedures. 1980) and comprise the perceptions employees have of their work environments with respect to safety policies. and measurement (Wright et al. and rewards (Griffin & Neal. & Hart. & Kelloway.14* Ϫ. Several elements of the high-performance work system result in trust in management.86) may be spuriously high owing to the possibility that companies who have systems of high-performance practices had both human resource and safety managers respond. namely. acts that go beyond employee roles.07 .. Employees’ perceptions of safety climate have been of interest for some time (see Keenan. If extended to the realm of safety. rather than simply to comply with external standards. employees who are trusting of management may be more inclined to work in a safe manner. Barling.. Although there does not appear to be any research on the effects of trust on safety performance.57%) warrants comment. 1999). which includes cooperation between individuals. For example. recent research suggests that low response rates may indeed not lead to biased findings. 1951. perceptions of safety climate will be more positive when workload is considered appropriate by employees (Zohar.43** Ϫ. but they do not address the question of how such an effect occurs. we extend the measurement of occupational safety. Griffin. In a meta-analysis of 86 articles in the occupationalhealth-psychology literature. management systems that promote employment security (Cascio. Individuals whose supervisors displayed safety-specific transformational leadership exhibited more positive perceptions of the safety climate and were less likely to engage in unsafe behavior (Barling et al. 1988. Becker and Huselid (1998) reported response rates for similar studies to be between 6 and 28%. (b) near misses—incidents at work that involve . more general evidence for the relationship exists. 1988.10 . restaurant. 2002). Similarly. Thus. 2002). we hypothesize that the relationship between human resource management practices and safety performance will be mediated by trust in management and safety climate. and perceived safety climate has predicted fewer injuries in production. 2000).. the response rate (9. Fifth. trust in management mediates the relationship between transformational leadership and employee well-being (Sivanathan. we accessed organizational records for the outcome measure. Although some caution needs to be expressed in terms of the implications. As we are unable to test this. In the prior study. BARLING.05.g.01. Zohar. 1993) and information sharing (Fitz-Enz. 1997) are likely to heighten employees’ trust in management. First and foremost. even though we supported the reliability of the lost-time-injury data.40** .08* c Note. 1999) and are invariably nonnormally distributed. Third. the correlation (r ϭ . 2001). 163). Kerr. Zohar. Zohar.33** . 1980. and young workers (Barling et al. 1 ϭ subsidiary. b 1 ϭ private. safety performance was only measured in terms of total lost-time injuries. although much of the prior research on high-performance practices has asked one person in each organization to provide the data on the practices. and to take greater initiative in safety-related matters. Study 2 In the second study.

2 respondents were female.OCCUPATIONAL SAFETY 85 Figure 1. All scales had five items with the exception of information sharing which had six. of schooling (SD ϭ 2. safety motivation. We conducted a pilot study in which a pool of 81 items were developed and tested. and they provide useful information regarding the actual state of safety in the organization.26). It was not possible to calculate the overall response rate in the petrochemical plant and one of the two telecommunications plant: Managers were responsible for distributing the surveys. SD ϭ 9.02). First. M years in current position ϭ 3.83%. Of the questionnaires that were completed. We also obtained data on personal-safety orientation. The majority were permanent employees (89. Second. Proposed model: Effects of a high-performance work system on occupational safety at the employee level. and we cannot know how many were distributed. In the other telecommunications site. The average number of hours of overtime worked was 3. and safety initiative. HPWS ϭ high-performance work system. Method Participants Participants in this study were 196 employees of two Canadian organizations from the petroleum and telecommunications industries.18) and had been in their positions an average of 7. Parker. Many items were similar between the two studies. in Study 2. Descriptive statistics and intercorrelations for the 10 practices appear in Table 4.77). Reason. the items used for this second study emphasized employees’ perceptions of the extent to which the organization had adopted the human resource practices. safety infractions but do not result in injury—and (c) lost-time injuries—the number of days of work lost because of injuries. a variable that comprised safety knowledge. Items were discarded on the basis of low item-total correlations and nonnormal distributions. Participants had worked in their respective organizations an average of 12. 142 responses were received (70 women. Responses were on a 5-point Likert scale ranging from strongly disagree (1) to strongly agree (5). SD ϭ 5.22 years (SD ϭ 7. item responses were all on Likert-type scales. Eightythree participants worked in the petroleum company as plant and field operators.68 hr per week (SD ϭ 8. compliance with safety rules and regulations. M age ϭ 32. and their surveys were removed to eliminate any effects that gender may have had on responses.84 years . Items measuring employment security were adopted from Kuhnert and Vance (1992) whereas items measuring transformational leadership were adopted from Bass and Avolio’s (1995) Multifactor Leadership Questionnaire. 72 men. while 113 worked in the telecommunications company as field technicians at two different sites.92) and had completed 13.12 years (SD ϭ 7. Participants (n ϭ 189 men) had a mean age of 39. other items were adapted to better suit the individual-level focus of Study 2. The items used to measure the high-performance work system in Study 2 differ somewhat from those used in Study 1 in two ways.60%). and Lawton (1998) suggested that these broader measures are particularly important because they provide the organization with continuous feedback—injuries requiring time away from work and fatalities are rare events whereas the personal-safety orientations of employees and the less severe safety incidents they experience are always present. Between 8 and 10 items were generated for each of the 10 practices.04).38 years (SD ϭ 9. with a response rate of 26.33.25). resulting in 51 items.39 years. 123 surveys were distributed and 33 returned.22 per week (SD ϭ 4. Most (84%) worked full-time for an average of 41. Materials High-performance work systems. Surveys were distributed via email to a sample of convenience of 349 individuals who were either recent graduates of an executive MBA program or their peers (n ϭ 225) or employees of a university’s Physical Plant Services Department (n ϭ 94).29. 191 contained usable data.

Hu and Bentler (1999) recommended that for samples below 250.001. Tanguma (2001) reported the CFI to be stable and less affected by sample size than other fit indices. 2002.. In relation to construct validity.51** . Browne. and Factor Loadings for the Pilot Study (N ϭ 142) Variable 1.4 Factor loadings (all significant and ranging from .56 for between sites.05. general item was also added to the scale.001.74 28. Turner.30** . P. F(2. the incidence of workplace injuries requiring first aid was assessed by asking individuals to think back over the past 6 months and to report how often (never ϭ 1 to frequently ϭ 5) they had sustained injuries requiring first aid at work in each of the same eight injury categories used in Study 1. * p Ͻ .001. which allowed the correlations between high-performance work systems and affective commitment to be constrained to unity (i. The corrected F test was nonsignificant. & Dusig. see Bliese.65 . 1999. As a result. 1992).5.78 . we calculated the difference between one model. & Gerras.55 . C. ** p Ͻ . Seven items derived from Neal et al. 2003. The measurement model displayed acceptable fit: GFI ϭ . 6 There is some debate in the literature as to whether “climate” should be considered an individual difference (e.. Williams. Neal et al.299. models. Neal et al. and systems and procedures (3 items. 1992). N ϭ 189) ϭ 718. Finkelstein.47** 3 4 5 6 7 8 9 10 Factor loading .55** . within and between analysis. & Yammarino. p Ͻ . Given these results. Nevitt & Hancock. p Ͻ .84) are presented in Table 4.01.35 4. A further. Therefore. Burke. 1996). the degree to which measures of constructs are distinct).23 21. Altmann. Castro. Hu and Bentler (1999) also preferred that the CFI be used for small sample sizes (N Յ 250).78 .42** 2 .45** . it is appropriate to consider safety climate as an individualdifference variable (Yammarino & Markham. Six items from Cook and Wall’s (1980) measure of interpersonal trust at work measured trust in management.79 . the E test (␩B/␩W) of . 7.09.36** . N ϭ 189) ϭ 93. In terms of the discriminant validity (i. as the two-factor model was found to fit the data significantly better than both the null.e. 2003) or group/organizational variable (e. Kirby. 2000.86 . and Parker (2000) measured safety compliance. p Ͼ . six items from the revised Meyer & Allen. following Hemingway and Smith (1999). scale). Morgeson.56 6.60** . p Ͻ .97 as compared with . current research (e.51** . 8.86 ZACHARATOS.10 may not be considered a good fit by some (e.61** .54** .73 . suggesting that there is not a significant difference in variation within and between the sites. 2002. 1998). . RMSEA ϭ . Responses were measured on a 7-point Likert-type scale ranging from strongly disagree (1) to strongly agree (7). Safety knowledge was measured with four items from Neal et al.. Safety initiative was measured with eight items from Turner and Parker (2004).69** . The CFA using LISREL VIII (Jo ¨ reskog & So ¨ rbom.27** . Young. indicating that the variation within the sites was greater than the variation between the sites. & Roberts.g. Responses to all four measures of personal-safety orientation took place on a 5-point Likert-type scale ranging from strongly disagree (1) to strongly agree (5). the degree of association between measures of a construct) of high-performance work system (10-item measure) and affective commitment was supported. Bollen.. 2002. Employment security Selective hiring Training Teams Status distinctions Information sharing Contingent compensation Transformational leadership Job quality Measurement M 29.58** .001 (Bagozzi & Yi.e. & Chen.60** . we examined whether there were significant differences between the three research sites.78 .e.84 . for procedures). training (4 items).74** . communication (5 items). training. 2000).20 34.60** .47 to .50** . MacCallum.38** . Sixteen items were used to measure four dimensions of individuals’ perceived safety climate: management values (4 items).92). N ϭ 189) ϭ 17. 3.29** .58 was practically significant (30o test).71 . Last. N ϭ 189) ϭ 1.. Baltes.82 . Hayduk & Glaser. we carried out an additional analysis with affective commitment (i.44** . 2000. 5.04 26. 2.68 .10.49** . 2000. 2003b).28** . Six months 4 Although the root-mean-square error of approximation (RMSEA) of . 1992) supported the aggregation of the summed scale scores of 10 practices.. in a Monte Carlo simulation. Conversely. First.31** .31 5. Second.54** . 1984. Kelloway..93.76 . (2000) and H. Dansereau. we undertook additional analyses (i.88. affirmed the discriminant validity of the high-performance work systems measure. as high-performance work systems are associated with high-commitment management.15 29.39** . The hypothesized one-factor model was a significantly better fit to the data than the null model. Lacost.54** . 2000) has raised concerns over RMSEA criterion and sample size sensitivity. ⌬␹2(17. AND IVERSON Table 4 Descriptive Statistics..64** .04 3.49** . we aggregated the data. which allowed the correlations between the constructs to be free. 10...47** .47 . CFI ϭ . Griffin & Neal. Hence. Parker. 2000.g. safety motivation was measured with four items from Neal et al.89.g. 2003a. & Sugawara. Huff.62. and safety systems.g.46** . The 2 chi-square difference test between the two models. 4.56** . 9. Because the eta (␩) value for within sites was .64** .79 5. and one–factor. The convergent validity (i. 1998.. ⌬␹difference (1.70 . Yammarino & Markham. Zohar.74 . 1984.e. 1997. 5 We also observed that the measure of high-performance work systems displayed convergent and discriminant validity with the mediator variable of trust in management as well. 171) ϭ .48 6.05. We argue that perceived safety climate is an individual variable that encompasses an individual’s perception of the management value.43** ..90 indicate a good fit of the data (Byrne. Paxton.71** .67** . as well as the general agreement that values for CFI exceeding . on the basis of sample-size considerations. p Ͻ . 6. Intercorrelations. Alutto.70 1 .46** .. in italics.96. Using one-way analysis of variance (ANOVA).43** . perfectly correlated) and the other.80 .19* .65** .e.75 . all 10 subscales were combined to form a single variable reflecting a high-performance work system (␣ ϭ .51** .85 5. To affirm our decision to operationalize safety climate as an individual difference variable. 1988).23 SD 4. In addition.82 Note. RMSEA be used with caution. Responses were on a 5-point Likert-type scale ranging from strongly disagree (1) to strongly agree (5).89 . ⌬␹2(10. Employees were asked to report how frequently they had almost sustained the particular type of injury over the past 6 months. Curran. ⌬␹2(1. we conceptualize perceived safety climate at the individual level. Safety incidents were assessed in three ways.5 Mediators. we inversed the traditional F test (Dansereau et al. the incidence of near misses was also measured using the same eight categories of injuries and the same response scale.6 Personal-safety orientation. Hofmann. Barling et al.91 5.71** .61** .28 27.66** .41 31.82 4.53 28. BARLING. Scale reliabilities (␣) can be found on the diagonal.66 26.29** . communication.

48** . . & Kelloway. Second. and safety incidents. Trust in management was treated as a singleindicator latent variable (e. Although mediated model.27** . Latent variables were also created to reflect the high-performance work system and safety-climate variables. in italics.30** Ϫ.22** . 1998). 5.11.92. and safety motivation were indicators of a single latent variable—personal-safety orientation. 9..08 Ϫ. N ϭ 189) ϭ 35. Standardized parameter estimates for the measurement model are presented in Table 6. 3. personal-safety orientation. The number of lost-time injuries reported were few in number (2.59* .53 1 . GFI ϭ . NFI ϭ . p Ͻ . ns. the original path between trust in management and safety incidents became nonsignificant. this last variable was dropped from further analyses. Barling.. and RMSEA ϭ .86. therefore.64 29. The proposed mediational model provided an excellent fit to the data.001.13 . Barling.10 3. 6. GFI ϭ . The proposed measurement model.31* Ϫ. normed fit index (NFI) ϭ . we examined measures of comparative fit (NFI.26* .96.70 17. the model was further modified to allow these two single-indicator variables to covary. Latent variables were created to reflect two types of safety outcomes.04 34.37** . Because of the high correlation between the trust and safetyclimate variables (r ϭ .96.99. however. personal-safety orientation.16* Ϫ. ns. Employees were also asked to report lost-time injuries and resulting days lost.16. & Amandus.70** — Note. safety climate. 10.38** Ϫ. ns. 1998): (a) the fully mediated model (Figure 1). As workplace injuries requiring first aid and near misses were averages of the eight injury categories and were considered cause and not effect indicators. safety knowledge.27 6. 2.53** . These two subscales were then used as observed variables reflecting the latent factor of high-performance work system. the incidence of actual injuries and near misses was used as an indicator of a single latent variable. and PNFI ϭ . p Ͻ .89.78 . N ϭ 189) ϭ 103.40. we found it necessary to estimate three nested structural models (Kelloway.96. N ϭ 189) ϭ 40. RMSEA ϭ .48** .70** . ␹2(28. Zacharatos.33** 2 . safety initiative.83 . To address this problem.53. and safety incidents) provided an acceptable fit to the data: ␹2(27.05. (b) a partially mediated model in which there are additional paths be- tween the high-performance work system and the outcome variables. 1994).72** .96. trust in management.01. and the chi-square-difference test) and the overall parsimony (PNFI) of each of the two models (Kelloway.99.69** .70.8% of employees sampled reported experiencing a single lost-time injury in the previous 6 months) and. namely.21 Ϫ. Bender.27** . HPWS Trust in management Safety climate Affective commitment Safety compliance Safety initiative Safety knowledge Safety motivation First aid Near miss M 153. Landen.05 Ϫ. Anderson and Gerbing’s (1988) two-stage modeling approach was followed.22** . The result of the chi-square-difference test. The nonmediated model did not provide a good fit to the data. GFI ϭ .90.04 1.05.63** .13 . we only added the path between high-performance work system and personal-safety orientation. ns.11 Ϫ.88 . AGFI ϭ . safety incidents.57 16. RMSEA ϭ . In terms of comparative fit. 1998). but in the opposite direction than expected.47** .16 9. adjusted goodnessof-fit (AGFI) ϭ . This was accomplished by conducting an odd– even split on the scales (e.01). both models provided acceptable fits to the data.g. 4. safety compliance.05.64.12 9. the fully mediated model must provide a better fit to the data than the nonmediated model and a more parsimonious fit to the data than the partially mediated model (Kelloway. and (c) a third nonmediated model that consists of the partially mediated model with the paths from trust in management and safety climate removed.42** . ␹2(30. and in Table 5 Descriptive Statistics and Intercorrelations Between Measures in Study 2 (N ϭ 189) Variable 1. ** p Ͻ . CFI ϭ .60. NFI ϭ .59 4.38 11. The partially mediated model also provided an excellent fit to the data. 1998). CFI ϭ .83.g.92 .27** . whereas the path between trust in management and personal-safety orientation became significant. CFI. HPWS ϭ high-performance work system.03 — .90 1. To determine whether the fully or partially mediated model provided a substantially better fit to the data.84.86 3. In the first model. RMSEA ϭ .54. p Ͻ . AGFI ϭ .16* Ϫ. The measurement structure was then used to estimate the structural relations of interest. To demonstrate mediation.91 . ␹2(32.31** .88 2. 2000). 1992). Kelloway. ␹2(29. we examined two partially mediated models.78 10.OCCUPATIONAL SAFETY is the recommended maximum time over which employees should be asked to recall injuries they have sustained with any accuracy (Veazie.64. All analyses were based on the covariance matrix and were estimated using maximum-likelihood estimation as implemented in LISREL VIII (Jo ¨ reskog & So ¨ rbom.. Scale reliabilities (␣) can be found on the diagonal. it was not appropriate to estimate coefficient alpha (Frone. These findings indicate a potential problem of overfitting the model (Kelloway. * p Ͻ .96. and PNFI ϭ . high-performance work system. 1998).95 .34 29.93.14 Ϫ.30** .99.55** Ϫ.52** . specifying five latent variables (i. 87 Results Descriptive statistics and intercorrelations of all variables in this study are presented in Table 5.35** . & Bremermann.92.84 .34** .68 SD 21.e. and parsimony normed fit index (PNFI) ϭ . 7.18 32. N ϭ 189) ϭ 43. GFI ϭ .13 Ϫ. In the first case. ␹difference the path between the high-performance work system and personalsafety orientation was significant. demonstrates that the partially mediated model was a significantly better fit than the fully 2 (2. 1991).42 64. The mediated model clearly provided a better fit to the data than the nonmediated model. To establish mediation. N ϭ 189) ϭ 6.04. CFI ϭ .89 .28** . CFI ϭ .11 .37** Ϫ.28** 3 4 5 6 7 8 9 10 .54** . N ϭ 189) ϭ 36. 8.15 .05.

ns). because of (a) potential problems of overfitting. 3. HPWS 1 HPWS 2 Trust Safety climate Compliance Initiative Knowledge Motivation First aid Near miss HPWS .95 .01).76. In addition. could account for these differential findings. First. 1995. **p Ͻ .59 .82 . we added the path between high-performance work system and safety incidents. A more plausible explanation. 1986): The context-specific nature of perceived safety climate.. ns. Although it is difficult to demonstrate conclusively that monomethod bias is absent from our results. safety climate. (b) the additional analyses. this study relied exclusively on self-report data. In contrast.19. the results of this second study again extend our understanding of the conceptualization of a high-performance work system. These models failed to show a significantly better fit than the fully 2 mediated model: that is. 1985). given that (a) organizational level measures of injuries may themselves be methodolog- Figure 2. 2001) and (b) the high-performance work practices were measured by a number of items each (Becker & Huselid. Therefore. N ϭ 189) ϭ 3. 6. personal-safety orientation.73 . respectively.01. 5. ns. BARLING.10. Discussion The results make several contributions to our understanding of the link between high-performance work systems and occupational safety.30. Standardized parameters for the fully mediated model. and safety incidents. the other. As shown.19) but not on personal-safety orientation. *p Ͻ . our focus on self-report data is justified.77. Second. Third. Our results suggest differential mediational effects for perceived safety climate and trust in management.88 ZACHARATOS. 9. Future research might focus on whether safetyspecific trust in management serves a more substantial mediating role. ␹difference (1. First. however. p Ͻ . p Ͻ .01).98 Trust . p Ͻ . Korsgaard & Roberson. is that this is an unfair comparison (Cooper & Richardson. ns.e.97 .90 Note. 8. HPWS ϭ high-performance work system. This second study is not without limitations.62. 2. . The construct validity of the measure was affirmed by CFA and provided further support that the 10 high-performance work practices are highly interrelated and form a single underlying factor.60) and displayed a practical improvement in fit (Widaman. ␹2(29. as opposed to the context-free nature of trust in management in this study.76) and safety incidents (␣ ϭ .05) and safety climate (␣ ϭ .59 .86.05. these results replicate the widespread effects of high-performance work systems on critical organizational outcomes. the fully mediated model was retained. Podsakoff & Organ. 4.87 .. 1986) and by the nonsignificant link between personal-safety orientation and trust in management.88 .84. p Ͻ . 10. the possibility is minimized by the significantly better incremental fit of the five latent variables (i. however.64 versus . Perceived safety climate mediated the relationships between the high-performance work system and both personal-safety orientation (␣ ϭ .01) as was safety climate (␣ ϭ . 7.01) but not by trust in management (␣ ϭ . Personal-safety orientation was predicted by safety climate (␣ ϭ . therefore. N ϭ 189) ϭ 41. and at least two factors may explain this discrepancy. It is possible that trust in management is relatively less important in this context.35). In addition. trust in management mediated the effects of the high-performance work system on safety incidents (␣ ϭ . The standardized parameter estimates for the model are presented in Figure 2. and (c) the better parsimony fit. 1998) and are. high-performance work system. p Ͻ . and 2 ␹difference(1. AND IVERSON Table 6 Standardized Parameter Estimates for the Measurement Model (N ϭ 189) Safety climate Personal safety orientation Safety incidents Variable 1. This result is particularly noteworthy given that (a) the scales were reliable (Wright et al.35. more meaningful than those of previous studies. Trust in management was predicted by high-performance work practices (␣ ϭ . examination of the PNFI indicated that the fully mediated model was more parsimonious than the original partially mediated model (. trust in management. these results extend our understanding of how high-performance work systems affect employee attitudes and behaviors. N ϭ 189) ϭ 1. safety incidents were predicted by both trust in management (␣ ϭ .

which showed that many of the management practices that are frequently applied to improve organizational performance (e. 411– 423. (1994). J. very little research has attempted to bring together these two distinct areas of study. the factor loadings of the subdimensions ranged from . Morgeson. Hofmann. 2000). this study confirms the role of organizational. J. job satisfaction and occupational safety. Arthur. The link between business strategy and industrial relations systems in American steel minimills. 17. C. Kelloway. 81. I. our data demonstrate that a high-performance work system is significantly associated with occupational safety. (1988). Zohar. The contribution of this work to our understanding in both these areas is significant. Second. and we therefore aggregated them into a single index measuring the high-performance work system. P. Journal of Applied Psychology. However. and supports the need to take a broader look at the understanding and management of occupational safety. (2000).. Kelloway. pay-based performance. J. & Yi. Our findings show the extensive benefits of high-performance work systems. 488 – 496. A further avenue for future research would be to examine these types of issues using a cross-level analysis to unravel individual and group-level factors that affect occupational safety (e. 88. Rather.. 1999) and (b) findings from our two studies using organizational records and self-report measures both demonstrate the role of high-performance work systems in occupational safety. Our analyses of three separate samples support the view that the measurement error of human resource practices is an important issue. Psychological Bulletin. we found that the 10 human resource practices formed a single factor. Barling.g. (c) high-performance work systems formed a unidimensional construct in three separate samples. J.g. With respect to occupational safety.. 2003b). 01–102). to date. no. through the use of CFA. 276 –283. In our analysis in Study 1. respectively (see Table 5). the latter two constructs were more strongly related to the high-performance work system than some of its indicators (in 3 out of 10 cases). & Kelloway. which could be differentiated from other constructs (e. Given the findings of these two studies. (2000). causal inference is not justified from the results of this study. R. teams) may have equal or greater effects on occupational injuries. formative vs. Nevertheless. 103. Weber. ergonomic design. Industrial and Labor Relations Review. References Anderson. E. factors in promoting safety. & Kelloway. Preemployment predictors of union attitudes: The role of family socialization and work beliefs. C. 37. W.70 and . we considered it to comprise several subdimensions. Work injuries and diseases (1997–1999). rather than individual. trust in management). whereas the composite measure of our high-performance work system was correlated at . An extension of the current findings in which control-oriented practices were also measured would allow researchers to make a direct comparison between the commitment-oriented and control-oriented practices and their impact on workplace safety. E. T. Barling. we attempted to demonstrate this construct validity. Journal of Applied Psychology. E. safety reputation.. these two studies have shown that (a) high-performance work systems affect occupational safety and (b) this relationship is mediated by trust in management and perceived safety climate. Barling. these findings have strong implications for interventions designed to enhance occupational safety.. Journal of Applied Psychology.’s (2003) findings showing that job satisfaction mediates the relationship between high-quality work and occupational safety. (1991). Barling. 621– 628. Canada: Author. On the evaluation of structural equation models. Barling. D.. 2003. (1992).. Mississauga. G. Bagozzi. W.. some caution needs to be expressed given the current concern regarding the causation between constructs and measures (i. First. Conceptually.84 (see Table 4). a further examination of other human resource practices that affect workplace safety would be timely. 488 –506. The mediational roles of job satisfaction and perceptions of organizational justice might also be examined as mediators in the high-performance work system–performance link.. Academy of Marketing Science.. the findings from these studies have important implications for the management of occupational safety. control-based safety practices. B. the consistency and reliability of reports of the highperformance work system warrant comment.69 with trust in management and affective commitment. Eisenberg & McDonald.. One remaining issue that deserves attention is the construct validity of the high-performance work system measure. . Development and test of a model linking transformational leadership and occupational safety. Empirically. K. especially given Barling et al... K. (1988). and longitudinal data are required to allow such inferences. 1988. in Study 2. (1996).. Jr.. 827– 832. High-quality jobs. General Discussion Several general issues emerge from these two studies that have conceptual. Pransky et al... Therefore. In addition. and researchers need to be attentive to the issue of withingroup consistency of these practices. Arthur.. (2003). Effects of human resource systems on manufacturing performance and turnover. 74 –94. and perceived safety climate. Effects of transformational leadership training on attitudinal and financial outcomes: A field experiment. B. K. & Hutchinson. (National work injuries statistics program. Third. & Gerras. Academy of Management Journal.47 to . 76. These findings are consistent with Kaminski’s (2001) findings. E. 76 – 84. Second. or practical implications. and government regulations of collective agreements.. Barrett. Research interest in the area of high-performance work systems continues to develop as does interest in the impact that management can have on workplace safety. Nevertheless. reflective. & Bremermann. (2001) have noted concerns about potential measurement problems in research on the effects of high-performance work systems. V. & Gerbing. E. Edwards & Bagozzi.OCCUPATIONAL SAFETY 89 ically flawed (e. J. 16. Structural equation modeling in practice: A review and recommended two-step approach. H. 75. Association of Workers’ Compensation Boards of Canada. Commitment vs.. (1990). R. 670 – 687. J.g. & Doverspike.e. Journal of Applied Psychology. Canadian Journal of Administrative Sciences. methodological. Journal of Applied Psychology. we can no longer assume that occupational safety is the primary prerogative of individual workers. 2003a.g. J. (2002). K. D. 725–731. Wright et al. Y. Loughlin. 45. Taken together. This highlights the difficulty and limitations of using multifaceted constructs such as high-performance work systems within a psychological model. Fourth. Cat. & Iverson. D. Arthur. 87. Validation of an information-processing-based test battery for the prediction of handling accidents among petroleum-product transport drivers. J.

J. (1998). & Wall. New York: AMACOM. 136 –162). Dyer. 54. (2000). P. Hopkins. 39 –52. DC: Author. J. 155–174. B. 1–36. The noncentral chi-square distribution in misspecified structural equation models: Finite sample results from a Monte Carlo simulation. H. T.. Journal of Applied Psychology. J. On the nature and direction of relationships betweens constructs and measures. 2. L.. Within-group agreement. DC: American Psychological Association. 6. H. L. & Chen.). Becker. (1999). (1986). Perceptions and worry about hazards at work: Unions.. Cascio. within. Dansereau. nonindependence. Colligan. Fenn. Cortina. 648 – 659. P. Predictors of work injuries among employed adolescents. 238 –246.. M. F. Newbury Park. 1999 (USDL Publication no. The impact of human resource management on organizational performance: Progress and prospects. A. (1998). J. S. S. 219 –233. K. 69 –93. L. Bass. Fitz-Enz. L. and motivation. A. Griffiths. Grunberg. Relations Industrielles/Industrial Relations. contract maintenance. Carlson. Leadership Quarterly. 78. The relationship of employee ownership and participation to workplace safety. K.. W. J.S. P. Paxton.. S. J. Journal of Applied Psychology. & McDonald. J. Bollen & J. Gillen. & Bagozzi. Long (Eds. (1985). 565–576. 16. and union density. Monthly Labor Review. Godard. Bielema.. D. S. 37(1). Common methodological problems in factor analytic studies. rwg(j). J. Data analytic methods for the analysis of multilevel questions: A comparison of intraclass correlation coefficients.. & Gilmore. (1993). Journal of Applied Psychology. 51–59. (2002). Barling & M. (2001). J. Conway. petrochemical industry. & Richardson. & Hoffman. Ergonomics. E. knowledge. 56. commitment. M.. New dawn or bad moon rising? Using large scale government administered workplace surveys and the future of Canadian IR research.. 46. 3–17. Journal of Organizational Behavior. L. A. NJ: Erlbaum. & Vaccaro. B. 00 –236).. W. B. Burke. Workplace risk. J. Drasgow.. (1993). M. M. & Neumark. & Anger. 20.. Moore. R. Frone. Kozlowski (Eds. 223–248). Palo Alto. Alternative ways of assessing model fit. B. San Francisco: Jossey-Bass. 656 – 670. B. R. Transformational leadership: Industry. 46. L. (1999). PRELIS. A. Griffin.. 53. M. Curran. The 8 practices of exceptional companies: How great organizations make the most of their human assets. & Yammarino. DeJoy. & Garber. & Ashby. and innovation. P. (1996). 111(4). 73. . Equivalence of psychological measurement in heterogeneous populations. Journal of Occupational Psychology. Human resource strategies and firm performance: What do we know and where do we need to go? International Journal of Human Resource Management. 58 – 60. H. B. 179 –184. F. 347– 358. Structural equation modeling with LISREL.. A. D. The role of social support in the stressor-strain relationship: An examination of work-family conflict. Journal of Applied Psychology. (1978). J.). 70. D. Journal of Applied Behavior Analysis. 27. D. Psychological Reports. Boredom at work: A neglected concept. 221–241. and reliability: Implications for data aggregation and analysis. 1067–1076. T. A.and between-analysis. (1993).. turnover. 107. Law and Policy. establishment size.. Economic and Industrial Democracy. & Reeves. R. J. Absenteeism and accidents in a dangerous environment: Empirical analysis of underground coal mines. M... A.. Journal of Occupational Health Psychology. Comrey.. 121(11). M. R. 15. Business responses to the regulations of health and safety in England. C. 75. M. L. Journal of Safety Research. L. R. (2002). 1– 8. (1995). (1996). D. W. extensions. & Gerhart. J. NJ: Erlbaum. (1980). & Davis-LaMastro. Washington. 33–51. 779 – 801. Genn. The long-term effects of a token economy on safety performance in open-pit mining.. Fox. 39. (1988). M. Occupational injury and illness rates. 98 –104. Baltz. W. P.. S. Kirch. D. Do “high-performance” work practices improve establishment-level outcomes? Industrial and Labor Relations Review. J. hierarchical linear modeling. Theory testing in organizational behavior: The variant approach. Castro. M. Geller. 25. Multilevel theory. H. Journal of Safety Research. (1988). Psychological Methods. New work attitude measures of trust. 49 – 68. M. CA: Mind Garden. M. The role of training in promoting workplace safety and health. (2001). What is coefficient alpha? An examination of theory and applications. (1999). 20. F. (1997). & Huselid. Multivariate Behavioral Research. Journal of Applied Psychology. R.. P. 42. 513–540. B. 38. AND IVERSON Cooper. Journal of Consulting and Clinical Psychology. Accidents. (1995). J. 17. R. Monthly Labor Review. J. Predicting propensity to actively care for occupational safety. Perceived organizational support and employee diligence. On average deviation indices for estimating interrater agreement. Frone (Eds. Mahwah. Journal of Safety Research. In J. and coworker support among union and nonunion injured construction workers. Bliese. B. Browne. and use of a preemployment screening inventory. M. (2004). 1992–96: Why they fell. Mahwah. Factors in successful occupational safety programs. (2000). applications. and new directions (pp. S. Research in Personnel and Human Resource Management. 73. F.. Klein & S. Goodman. D. NJ: Prentice Hall. (1993). and programming. (1998). 33. J. Managing safety in the workplace: An attribution theory analysis and model. Englewood Cliffs. CA: Sage. Journal of Management. (2002). & Kanfer. Human Relations. 737–775. Perceptions of organizational safety: Implications for the development of safety culture. & Neal. 71. 25. 522–541.90 ZACHARATOS. High performance work systems and firm performance: A synthesis of research and managerial implications. P. & Svenson. Downsizing: What do we know? What have we learned? Academy of Management Executive. (2000). Fisher. Fasolo. A. K.. Journal of Applied Psychology.. Industrial Relations. Bentler.. Journal of Safety Research. Alutto. 28. and methods in organizations: Foundations.. & Greenberg.. (1996).. S. Eisenberg. 61– 67. Psychological Bulletin. (1987). (1998). Cappelli. Perceived safety climate.. M. In K. S. 36 –58. S. BARLING. 395– 417. and job control in the U. 95–104. (1990). 168 –178. 9. 81– 86. K. 215–224. V. Academy of Management Journal... Becker.. Unfair comparisons. A. (1998). Cook. 349 – 381). (1984). (1985). The psychology of workplace safety (pp. (1994). M. (2003). P. Multifactor leadership questionnaire for research. & Dusig. and random group resampling. M. Clarke. M. Baugher.. Roberts. 83. R. G. Finkelstein. D. 3–33. E. F (1993). Eisenberger. W. Byrne. Evaluating workplace injury and illness records: Testing a procedure. Comparative fit indexes in structural models.. 662– 680. D. & Roberts. 53–101. L. & Cudeck. 13. Testing structural equation models (pp. (1999). 5. & Perrewe ` . P (2000). Perceptions of safety at work: A framework for linking safety climate to safety performance. Cohen. military and educational impact.. organizational commitment and personal need non-fulfillment. research. 461– 480. J. E. 7. M. Bass. 5. National census of fatal occupational injuries. W.). Washington. and SIMPLIS: Basic concepts. Edwards. British Journal of Industrial Relations. (1990). Borofsky. S. M. & Cohen. Bureau of Labor Statistics. Safety attitudes of management. P. In K. S. Kirby. T.. Bollen. Gassel. D. job demands. & Avolio. Organizational Research Methods. A. (1977). (1993). 185–198. J.

training. productivity. Parker. L. Korsgaard. (1997). Osterman. Toronto. Psychological climate and accidents in an automotive plant. DC: American Psychological Association. 21. DC: American Psychological Association. J. 26 –39.. R. E. K. (1992). J... & Gerras. J. Huselid. In J. 35.. 49. A. P. A. 1–55. 48. M.. (1995). Journal of Occupational and Organizational Psychology. The causes of coal mine accidents. R. & Stetzer. 35. & Salas. (1997). B. H. J. Jones. Lacost. Thousand Oaks. turnover and productivity: Evidence from New Zealand. M. P. G. Kramer. Monthly Labor Review. . productivity. August). (1995). LISREL VIII: Analysis of linear structural relations. & Morgeson. G. 251–268.). R. Trust and distrust in organizations: Emerging perspectives. R. Job insecurity and moderators of the relation between job insecurity and employee adjustment. Labor unions and occupational safety: Conflict and cooperation. G. Baltes. J. (1998). J. (2003). R. Paper presented at the Academy of Management Annual Meeting. 407– 416. A cross-level investigation of factors influencing unsafe behaviors and accidents. & Erwin. (1998). J. G. Jo ¨ reskog. L. B. (1996). waltzing around factor analysis. (1996). B. Safety leadership in the offshore oil and gas industry. M. B.. Hopkins.. E. C. In J.. M. D. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives.. D. Academy of Management Journal. Hurrell. A. R. (2000).). J. Washington. H. P. Academy of Management Journal. W. G. Improving the root mean square error of approximation for nonnormal conditions in structural equation modeling. 35. Kochan. M. Hemingway. A. Work redesign. and other serious fun. Commitment in the workplace: Theory. 6. Safety-related behavior as a social exchange: The role of perceived organizational support and leader–member exchange. Martell. Hu. Journal of Applied Psychology. P.. 291–313. Barling & M. 307–339. (1999). 7–19. Predicting occupational injury: The role of affectivity... (2001). Murphy. A.. & Palser. (1997). MacCallum.. 44. Structural Equation Modeling. K. R.. 20. E.. 34. Opening the black box: An experimental investigation of the mediating effects of trust and value congruence on transformational and transactional leadership. Journal of Organizational Behavior. M. A. 1. Journal of Occupational Health Psychology. enduring questions. span of control and safety performance in work teams after workforce reduction. Ichniowski. Methodological issues in crosssectional and panel estimates of the human resource-firm performance link. Industrial Relations. 185–194). (1982). Hofmann. Job demands. The psychology of workplace safety. A. J. & Becker. & Flin. & Salminen.. Industrial and Labor Relations Review. 2. 29 –36. Industrial Relations Journal. and appraisals of management: A study of nuclear power plant personnel. P. (1996). What works at work: Overview and assessment. (1999). 38. Ichniowski. Haynes. Journal of Occupational Health Psychology.. (1995). Olson. (1991). C. 170 –178. Journal of Applied Psychology. High performance work systems and firm performance. Hackman. Young. 34. The impact of human resource practices on turnover. Canada. From the ground up: Six principles for building the new logic corporation. Industrial Relations. Empowerment. 569 –598.. (2000). Relationships between psychological climate perceptions and work outcomes: A meta-analytic review. D.). E. Journal of Experimental Education. 25. Kivima ¨ ki. O’Dea. D. J. R. 84.. M. & Oldham. A. Administrative Science Quarterly. A. 88. (1995). & J. N. W. A.. Unintended consequences: Organizational practices and their impact on workplace safety and productivity. J. (2001). and mental strain: Implications for job redesign.OCCUPATIONAL SAFETY Guthrie. Altmann. M... 253–267. P. 400 – 422. A. McAllister. How strategic is HRM? Human Resource Management. MA: Addison-Wesley. R. 645– 672. (1995). (1990). Journal of Organizational Behavior. & Sugawara. Hayduk. Academy of Management Journal. I. Shaw. & Carroll. Journal of Applied Psychology. C. A. 118(5).. Huff. A. R. & Winn. D. (1951). Iverson.. 24. & Hancock. (1979). K. (1995).. 70. Reducing accident rates with organizational behavior modification. R. E. C. V. organizational commitment. M. B. High involvement work practices. J. (2000. D.. 72. C. D.. 108 –111. and work organization in American establishments. R. J. J. Organizational climate and occupational stressors as predictors of withdrawal behaviors and injuries in nurses. 99 –109.. W. 285–299. S. & Roberson. C. (1995). Hofmann. Meyer. K. Keenan. T. Preemployment honesty testing (pp. 48 – 63). Driskell. Frone (Eds. 130 –149. 91 Karasek. & Roberts. P. F. Kelloway. & Hart. J. Using LISREL for structural equation modeling. Reading. M. M. F. 34. New York: Quorum.. In J. Journal of Management. D. 6. S. Journal of Safety Research. A. 18. Annual Review of Psychology. W.. & Allen. 389 – 416. 21. L. The impact of organizational climate on safety climate and individual behavior. A. N. job decision latitude. P. D. 275–282. Climate as a moderator of the relationship between leader–member exchange and content specific citizenship: Safety climate as an exemplar. & Beehr. S. Human resource bundles and manufacturing performance: Organizational logic and flexible production systems in the world automotive industry. E. New York: Basic Books. 949 –964. S. Psychological Methods. Kaminski. 24. 50. S. L. 113–128. (2001). Hechanova-Alampay. Industrial Relations. 657– 669. J. Thousand Oaks. Kling. CA: Sage. (2000). T.. H. (1999). H. E. & Prennushi. Research and Practice. & Sherman. American Economic Review. D. 87. R. Hofmann.. Browne. Morgeson. The use of incentives/feedback to enhance work place safety: A critique of the literature. J. and the reconstruction of working life.. Skill. 15. & So ¨ rbom.. W. Neal. A. W. A. R.. 299 –333. & Smith. Group Dynamics: Theory. & Strauss. G. Power analysis and determination of sample size for covariance structure modeling.. Washington. 180 –190. A. IN: Scientific Software. (1980). research and application. A. R. Kalimo.. 391–396.. J. III. P. Jung. E. Safety Science. T. Huselid. Nevitt. M. Healthy work: Stress. & Theorell. 6. Stress and well-being at work: Assessments and interventions for occupational mental health (pp. K. San Francisco: Jossey-Bass. Griffin. A personnel selection approach to industrial safety. (1999). 285–308. Perceived nuclear risk.. (2003). Lawler. Kelloway. (1992). 1–35. Status and cockpit dynamics: A review and empirical study. (1989). 7.. CA: Sage. Journal of Occupational and Organizational Psychology.. Pine. Milanovich. & Glaser. Quick. MacDuffie. The effects of human resource management practices on productivity: A study of steel finishing lines. A. J. Jr. & Fitch. R. (2000). Structural Equation Modeling. (Eds. Procedural justice in performance evaluation: The role of instrumental voice in performance appraisal discussions. Personnel Psychology. Stout. (1987)..and cognition-based trust as foundations for interpersonal cooperation in organizations. W. 38. (1996). Risk Analysis. C.. and corporate financial performance. 197–221. P. 68. Levine. Jiving the four-step. R. Jones (Ed. Academy of Management Journal. Mooresville. 286 –296. Affect. J. M. Kerr.. C.. K. R. & Bentler. Kuhnert. (1996). McAfee. S. & Vance. A. 24 –59. & Avolio. (1995). R. 155–167. 125–146. (2003)... 127–138. Karasek.

E. (1995). J. 8. (2003). A. Park. A (2002). (2001). 76. & Marchand. Yammarino. T. Employees and high performance work systems: Testing inside the black box. S. H. M. Meeting trainees’ expectations: The influence of training fulfillment on the development of commitment. Parker.... F. & Wall.). Cohen. command and control. Tsui. E. Can we speak of a high commitment management on the shop floor? Journal of Management Studies. 177–191. Journal of Applied Psychology. Journal of Management. Toronto. DC: Author. N. (1985). G. Warr (Ed. Testing structural equation models (pp. W. 351– 374). Warr (Ed. Simard. (1985). Behavior based safety programs. London: Penguin Books.. (2003). (1997). I. (1998). Academy of Management Journal. L. S. (1998). Williams. T. & Parker. Thousand Oaks. E. 171–182. Canada. who is Pied Piper and who are the rats?” Paper presented at the Association of Workers’ Compensation Boards of Canada Congress. M. J. D. H. S.). (2004). “That’s not my job”: Developing flexible employee work orientations. Journal of Management. Warr. M. & Marchand. 875–901.. D. and company performance. Gerhart. Tanguma. 113–129. J. H. Williams. Industrial Relations. & Delery. 25. Human Relations. E.. W. Loughlin. 10. 40. (1999). 42. Moynihan. From control to commitment in the work place.. 201–236. Do “high commitment” human resource practices affect employee commitment? A cross-level analysis using hierarchical linear modeling. T. R. E.. J. 258 –268.. Probst. pp. S.. 391– 417. Designing a safer workplace: Importance of job autonomy. 15. Professional Safety. and motivation. Goodbye. 12. S. A. 515–535. White. Pearce. J. M. E. T. Moran. (1986). & Lasaosa. 40. V.. (1998. (1997). Integrated manufacturing. H. Effects of sample size on the distribution of selected fit indices: A graphic approach.. Journal of Occupational Health Psychology. D.). Barling.. Job and work design: Organizing work to promote well being and effectiveness. 54. The human equation: Building profits by putting people first. Walton. 71. C. Newbury Park. & Jackson. The effects of job insecurity on employee safety outcomes: Cross-sectional and longitudinal explorations. K. P. Oxford. 289 –304. T. S. Journal of Occupational Health Psychology. Antecedents and consequences of satisfaction and commitment in turnover models: A reanalysis using latent variable structural equation methods. B. I.. Family-friendly management in Great Britain: Testing various perspectives. Safety Science. B. K. Dembe. K. J. 9. 40. (1987). 35– 62). (1991). M. S. J. N. 501–531.. 24. Manuscript submitted for publication.. 38. Axtell. Group and Organization Studies. Haines. U. Landen. L. Psychology at work (5th ed. L. L.92 ZACHARATOS.. (2001). L. T. N. L. 77– 84. K. J. C. 37. American Journal of Industrial Medicine. 203–221. J. S. (2002). (1997). Work. (1986). Patterson. & Parker. E. (1978). P.. and supportive supervisors. Harvard Business Review.S. B. A. S. (1997. Journal of Organizational Behavior. T. Salas. AND IVERSON Tjosvold. Wood. J. 63(2). D. M.. Mathieu.. S. & Cannon-Bowers. 42. (1999). 139 –159. Wood. K. A. Leading others to well being: Transformational leadership and employee well-being... R. A multilevel analysis of organizational factors related to the taking of safety initiative by work groups. Podsakoff. 1089 –1121. B.... 28. Wood. Hierarchically nested covariance structure models for multitrait-multimethod data. Journal of Safety Research. 37–39. S. (1999). J. M. H.. Annual Review of Public Health. T. Wheatley. G. Alternative approaches to the employee-organization relationship: Does investment in employees pay off? Academy of Management Journal. & Kelloway. Whitener. G. J.. 23–29. Cohen. R. Shannon. Turner. & Amandus. & Hazer.. 219 –231. (1977). & Hodson. T. BARLING. & Wall. L. & Wall. Trist. A. J. Journal of Occupational and Organizational Psychology. (1992). Underreporting of work-related disorders in the workplace: A case study and review of the literature.. & Organ. S. Walker. M. Canada.. Educational and Psychological Measurement. Widaman. (1998b).. Porter. D. CA: Sage. In P. August). 221–250. Winnipeg. E.. Workplace organizational correlates of lost-time accident rates in manufacturing. unemployment. Ergonomics. L. Washington. C. D. and mental health. In K. British Journal of Industrial Relations.. Leader to Leader. Psychology at work (pp.. & Brubaker. K. S. B. Getting the measure of the transformed high-performance organization.. (2001). (1997).. (1990). Flight crew collaboration to manage safety risks. 765–785. Wood. Self-reports in organizational research: Problems and prospects.. (1993). & Himmelstein. The effect of teamwork on safety processes and outcomes. M. Washington. Ramsay. (1994). Journal of Applied Psychology. de Menezes. D. K. CA: Sage. M.. J. M. 6. 172–188. Frone (Eds. Workgroups’ propensity to comply with safety rules: The influence of micro-macro organizational factors. S. Wood. 759 –776. (2001).. Summer). J.. 71. 21–28. Power evaluations in structural equation models. On the application of within Parker.. The compensatory role of transformational leadership in promoting safety behaviors. D. Snyder. 6. 215–247. July). G. (1998a). & Verma. Turner. R. R. Human resource management and business performance. England: Oxford University Press. Smith. N. Simard. self-efficacy. Ergonomics. Sprigener. & Harley. M.. J. S. T. (2000. E. 1–26. & Brown.S. An experiment in autonomous working in an American underground coal mine. D. & Turner. F. Parker. T. J.. D. Manitoba. T. 15.. Pfeffer. 30.. & Cleveland. A. 27. M. & Lawton. British Journal of Industrial Relations. Kraus. Tannenbaum. A.. Census Bureau. E. (1995). Wright. Schalm. Richardson. I. Epidemiologic research on the etiology of injuries at work. A. H. P. High performance work systems and intermediate indicators of firm performance within the U. (2001). Statistical abstract of the United States: 2000 (120th ed. In J. S. Barling & M. Organizational controls and safety: The varieties of rule-related behavior. Susman. R. R.). & Markham. & Wall. 32. 181–204). & Tripoli. Personnel Psychology. A. L. 759 –769. H. empowerment. Pfeffer. Reason. Journal of Management. Boston: Harvard Business School Press. Long (Eds.. Characteristics of successful safety programs. Lewchuk.. Adopting high-involvement human resources practices. or “if it’s rat psychology. & Satorra. London: Penguin. J. & Albanese. West. 5. S.. 641– 665. K.. 899 –929. 1–16).. In P. Veazie. Scholarios.. & Kelloway. E. 160 –163. small business sector. P. 5–15. Parker.. A. 29. 531–544. J. (2004). DC: American Psychological Association.... (2000). (1996). Bender. (2001). Sivanathan N. A. D. S. Pransky. The psychology of workplace safety (pp. Paper presented at the Academy of Management Annual Meeting. Way. Group & Organization Management. Gardner.. Applied Psychological Measurement. Walters. Wall. Leader to Leader. 6. 21. The relationship between response rate and effect size in occupational health psychology research. Human resource management and business performance. Are labor unions in danger of losing their leadership position in safety. communication quality.. . D. (2001). (2000). E. A Bollen and J. & Williams. Journal of Occupational Health Psychology.. M. M. J. The real keys to high performance. Sanchez. T. 461– 478. 42.). Measurement error in research on human performance: Additional data and suggestions for future research.. 61. 211–228. Saris. R. W.... A.

D. A. Hoffman & L. 2003 Revision received January 26. The effects of leadership dimensions. Barling. Tetrick (Eds. Zohar. K. 168 –176. 77. C. (2003a). E. Tetrick (Eds. Safety climate in industrial organizations: Theoretical and applied implications. Zohar. Zohar. A group-level model of safety climate: Testing the effect of group climate on microaccidents in manufacturing jobs. Quick & L. 123–142). E. J. 211–226. 2004 Ⅲ . Received January 20. E. 85.. 201–230). 23. D. In D. A. (2002). Zacharatos. 75–92. Journal of Applied Psychology. 65. Zohar. 2004 Accepted January 28. Leadership Quarterly.). safety climate and assigned priorities on minor injuries in work groups. DC: American Psychological Association. San Francisco: Wiley.OCCUPATIONAL SAFETY and between analysis: Are absence and affect really group-based phenomena? Journal of Applied Psychology. Handbook of occupational health psychology (pp. Journal of Organizational Behavior. & Kelloway. (2000). 96 –102. 587–596. (1980). 11. The influence of leadership and climate on occupational health and safety.). In J. (2003b). (2000). D. D.. Washington. Health and safety in organizations: A multilevel perspective (pp. Journal of Applied Psychology. Development and effects of transformational leadership in adolescents. Safety climate: Conceptual and measurement issues. D. 93 Zohar.